Fig. 2: GbM and teM explain substantial amounts of gene expression variance.

a, Density plots grouping successfully modelled genes (N = 7,339) by the proportion of the expression variation explained by genome-wide gbM (gold), genome-wide teM (blue), SNPs (green) or all three (brown). b–d, Genes were split by their population gbM frequency (<10%, N = 1,970; ≥10 and <90%, N = 1,508; ≥90 and <100%, N = 1,970; 100%, N = 1,891). The proportion of expression variation explained by SNPs (b), gbM (c) and teM (d) is plotted. e,f, Genes were split by Shannon entropy of expression (<3.6, N = 1,766; 3.6–4.4, N = 3,584; ≥4.4, N = 1,935), and the proportion of expression variation explained by teM (e) and gbM (f) is plotted. g,h, GbM genes (gbM population frequency ≥90%) were split by Shannon entropy (<3.6, N = 427; 3.6–4.4, N = 1,854; ≥4.4, N = 1,564) and the proportion of expression variation explained by teM (g) and gbM (h) is plotted. Superscript letters after mean values in all panels signify P < 0.01 using the non-parametric Kruskal–Wallis test followed by pairwise comparisons using the Wilcoxon rank-sum test with Bonferroni correction for multiple testing. Groups sharing the same letter are not significantly different.